• Previous message: Lynn Hathaway: "imagine if you had a diploma yrnfirry rcpmqson"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

number9 AT plate DOT dimlight.org writes:
 > 
 > The techfile is SCN6M_SUBM.10 from the "current.tar" techfile
 > set. The version of magic is 7.1
 > 
 > All I do is start magic and:
 > 
 > :getcell sram6t
 > :array 8 64
 > 
 > then I put some labels on it and do a:
 > 
 > :cif write sram6t_512b.cif
 > 

Hi Robert,

well with your original file set

  sram6t.mag       sramtestcif.mag 

(that i got back from our IT dept. who thought it was a windows virus)

using 'magic7.1 -T SCN6M_SUBM.10 sramtestcif'

i do get that cifout error:

:cif
CIF warning for array sram6t.mag_0 (instances of sram6t.mag):
    cells are so small and close together that two
    non-adjacent cells could have interacting CIF!

but i found that simply reducing the 'shrink 500' to 'shrink 190' (and
lower) in

cifoutput
...
 layer COG pad
 shrink 500
 labels pad
 calma 52 0

removes the problem, but that even 'shrink 200' still exhibits it.
i don't kow why those values (190 and lower ok, 200 and higher bad)
have that effect, given your cell sizes.  it's specific to arrays,
since individual cell placements have no problem.

(the size of the test cell is:

EDITbox width   height  llx     lly     urx     ury
        39      67      2       -1      41      66

and the 'array 2 1' has size:

EDITbox width   height  llx     lly     urx     ury
        78      67      2       -1      80      66
)

but the 500 is there to generate vias around the edge of the special
symbolic 'pad' layer (old style vias).

i agree that in this case, i don't think it's a real problem, but in
general you should never allow this cifout error to happen (the cifout
error may also be complaining about a real problem).

3 solns:

1. don't use array, rather place each cell with a macro (i think you
   discussed this method before).

2. make an overlapping 'even' array and 'odd' array.  that is, rather
   than

:getc sram6t
:array 4 1
:cif

you'd do:

:getc sram6t
:bo
EDITbox width   height  llx     lly     urx     ury
        39      67      2       -1      41      66
:box wid 78
:arr 2 1
Selected cell is sram6t (sram6t_0[0])
:bo ri
:getc sram6t
:bo wi 78
:arr 2 1
:cif

3. remove the 'pad' layer form the techfile (but a non-trivial edit).
then this cifout section to generate vias from 'pad' would be removed.


good luck!

/jeff

• Previous message: Lynn Hathaway: "imagine if you had a diploma yrnfirry rcpmqson"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]